Abstract
Vector multiplication involving multiple matrices is theoretically realized by a crossbar array of conventional phase-change random-access memories whose device resistance can be continuously adjusted by altering the amorphization/crystallization pulse numbers and amplitudes. The designed circuits comprise input current, memory conductance, and output voltage, to represent the known vector, matrix coefficient, and required solution, respectively. The calculated results from simulations show good agreement with analytical solutions, demonstrating capability for in-memory computation.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have